Z.ai: glm-ocr
10 benchmarks
Leaderboard position
#541
Overall rank (of 1121 models)
1013
Elo rating
10
Benchmarks
Top 48.3%
Percentile
In Z.ai (of 18 models)
Above: NousResearch: hermes-4-405b #540
Below: Google: gemini-2.0-flash #542
Benchmark highlights
Best benchmark
UniMERNet
96.5
Worst benchmark
ParseBench
29.6
Nearest competitor from another company
The model from another company with the most similar benchmark scores.
appeared as nearest in 4 of 10 benchmarks
4/10
40.0%
| Model | Votes | % of benchmarks |
|---|---|---|
| Google: Gemini 3 Pro Preview | 4 | 40.0% |
| OpenAI: gpt-5.2 | 2 | 20.0% |
| Qwen: Qwen3.5 0.8B | 1 | 10.0% |
| opendatalab: mineru2.5-2509-1.2b | 1 | 10.0% |
| Qwen: qwen3-vl-8b-instruct | 1 | 10.0% |
| PaddlePaddle: PaddleOCR-VL-1.5 | 1 | 10.0% |
Benchmarks
Predicted scores
Estimated from 25 similar models and the global Elo rating. These are projections, not measured results.
| Benchmark | Predicted | Confidence | Based on |
|---|---|---|---|
| MATH-500 | ~89.14 | 50% | — |
| AIME 2026 | ~76.56 | 50% | — |
| HMMT Feb 2026 | ~68.08 | 50% | — |
| GPQA Diamond | ~61.57 | 50% | — |
| MMLU-Pro | ~65.11 | 50% | — |
| Humanity's Last Exam (no tools) | ~14.46 | 50% | — |
| IFEval | ~84.44 | 50% | — |
| IFBench | ~55.96 | 50% | — |
| HumanEval | ~78.31 | 50% | — |
| MBPP | ~83.07 | 50% | — |
| Terminal-Bench v2.1 | ~27 | 50% | — |
| MMLU | ~71.76 | 50% | — |
| SWE-bench Verified | ~44.17 | 50% | — |
| SWE-bench Pro | ~22.78 | 50% | — |
| SciCode | ~29.4 | 50% | — |
| AA-LCR | ~46.2 | 50% | — |
| Artificial Analysis Intelligence Index v5.0.1 | ~12 | 50% | — |
| Humanity's Last Exam (with tools) | ~19.62 | 50% | — |
| GSM8K | ~88.78 | 50% | — |
| -(CoderPPL) | ~-1.69 | 50% | — |
| LiveCodeBench v6 | ~59.55 | 50% | — |
| MMMU-Pro | ~55.93 | 50% | — |
| MATH | ~67.42 | 50% | — |
| BBH | ~68.85 | 50% | — |
| ARC-Easy | ~89.9 | 50% | — |